Path::file_stem() does not work with directories that contain dots

[hacknus]

I'm dealing with directories that contain dots in the name:

let path = Path::new("2024.11.23_directory");
dbg!(path.file_stem());

I expected it to just return the pure filename, since it does not have an extension/suffix.
"2024.11.23_directory"

Instead, it just returns whatever is before the last dot:
"2024.11"

Meta

rustc --version --verbose:

rustc 1.82.0 (f6e511eec 2024-10-15)
binary: rustc
commit-hash: f6e511eec7342f59a25f7c0534f1dbea00d01b14
commit-date: 2024-10-15
host: aarch64-apple-darwin
release: 1.82.0
LLVM version: 19.1.1

This happens in the source code in path.rs where:

   pub fn file_stem(&self) -> Option<&OsStr> {
        self.file_name().map(rsplit_file_at_dot).and_then(|(before, after)| before.or(after))
    }

This clearly shows that it just splits the filename at the last dot, regardless if it is a directory or a file.

[ryanavella]

This is likely (?) intentional behavior. Some devs operate on abstract paths that don't necessarily exist on a mounted filesystem. It is also worth noting that checking if a path is a directory would introduce a new syscall to Path::file_stem.

[hacknus]

@ryanavella I disagree. A directory does not have an extension, so we cannot return a file stem (or extension for that matter). I see that an additional syscall is required for my suggestion, but I don't see the use of the file_stem() function if it does not handle that.

For abstract paths, which do not exist on the filesystem, the is_dir() method returns false, so the behavior wouldn't change.

[ryanavella]

For abstract paths, which do not exist on the filesystem, the is_dir() method returns false, so the behavior wouldn't change.

The problem is if a process calls Path::file_stem on an abstract path, and another process has coincidentally created a directory with the same name so that it now exists. For these abstract paths, Path::file_stem on the same Path could return different results depending on filesystem state.

[rvolgers]

Some systems do use extensions with directories. For example, on OSX, apps are distributed as app bundles, which are just directories with the extension ".app".

[hacknus]

Some systems do use extensions with directories. For example, on OSX, apps are distributed as app bundles, which are just directories with the extension ".app".

Fair point. Yes.

For abstract paths, which do not exist on the filesystem, the is_dir() method returns false, so the behavior wouldn't change.

The problem is if a process calls Path::file_stem on an abstract path, and another process has coincidentally created a directory with the same name so that it now exists. For these abstract paths, Path::file_stem on the same Path could return different results depending on filesystem state.

I see..

Python's pathlib also handles it the same way.

However, I still think that cases where there are dots occurring in the name of a directory, the behavior of the std::path library is not useable. But a solution is not as straight forward as I thought..